Company

Absci

AI drug-creation company using zero-shot generative models to design de novo antibodies, validated by an in-house wet-lab feedback loop.

1. Core Product / Service

Absci is a "data-first generative AI drug creation" company. Its Integrated Drug Creation™ platform couples generative AI models with a synthetic-biology data engine in a continuous loop: AI proposes antibody candidates, a high-throughput wet lab screens them (reported capacity of billions of cells per week), and the results feed back to retrain the models — going from AI-designed candidate to wet-lab-validated candidate in as little as six weeks [1].

Its flagship model, IgDesign1, is described as the first in vitro-validated inverse folding model for antibody design [2]. Absci also advances its own pipeline: ABS-101 (an IBD antibody in IND-enabling studies) and ABS-201 (hair regrowth / androgenetic alopecia) [1].

2. Target Users & Pain Points

  • Pharma / biotech partners who want de novo antibodies against hard targets without screening millions-to-billions of natural candidates.
  • Its own pipeline — Absci is both a platform vendor and a drug developer, using internal programs to prove the "AI-designed antibody" narrative.

Pain solved: traditional antibody discovery is a high-cost, low-throughput screening exercise; Absci compresses discovery into a computation-first loop with fast wet-lab validation [1].

3. Competitive Landscape

Player Approach Vs. Absci
nabla-bio JAM multimodal generative design + wet lab Both do de novo antibody design; Nabla is partnership-funded, Absci is public
chai-discovery Chai-1 structure / Chai-2 antibody design Chai is open-source-first with a ~20% hit-rate claim
evolutionary-scale ESM3 sequence/structure/function model Platform model licensing vs. integrated drug-creation loop
isomorphic-labs AlphaFold-lineage, Alphabet-funded Big-tech scale; Absci is an independent public biotech

Absci's differentiation is the tight AI↔wet-lab feedback loop and its own clinical pipeline, rather than selling model access alone.

4. Unique Observations

  • The binder speed/accuracy benchmark: this week's research compared Absci's de novo binder generation against an internal "Claude Science" agent pipeline on speed and accuracy — a live test of whether a frontier LLM agent can match a purpose-built antibody-design stack [local]. The "pollution" problem in de novo antibody design (proving generated binders are genuinely novel, not memorized from training data) is a shared open question for both.
  • Compute partnership as a signal: Absci's AMD collaboration (deploying Instinct accelerators + ROCm for antibody models) and its data-engine moat suggest the real barrier in AI drug design is proprietary wet-lab data, not raw model scale [2].

5. Financials / Funding

  • IPO: July 2021 on Nasdaq (ticker ABSI), priced $16/share, ~$230M gross proceeds, ~$2B debut valuation [3].
  • AMD strategic investment: $20M PIPE (Jan 2025) alongside the collaboration [2].
  • July 2025 raise: ~$64M gross ($50M underwritten offering + ~$14M ATM) [1].
  • Cash: $117.5M as of June 30, 2025; runway guided into H1 2028 [1].
  • Revenue: Q2 2025 $0.6M; FY2024 $4.5M; FY2024 net loss $103.1M [1].

6. People & Relationships

Sources

  • [1] Nasdaq, "Absci Reports Business Updates and Second Quarter 2025 Financial and Operating Results" (2026-08-24)
  • [2] Absci–AMD collaboration and strategic investment announcement (2026-08-24)
  • [3] GeekWire, "Absci shares spike 30% in IPO debut…" (2021) (2026-08-24)
Last compiled: 2026-08-24